CN105843857A - Video recommendation method and device - Google Patents
Video recommendation method and device Download PDFInfo
- Publication number
- CN105843857A CN105843857A CN201610151232.6A CN201610151232A CN105843857A CN 105843857 A CN105843857 A CN 105843857A CN 201610151232 A CN201610151232 A CN 201610151232A CN 105843857 A CN105843857 A CN 105843857A
- Authority
- CN
- China
- Prior art keywords
- performer
- video
- weight
- target video
- unit
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/70—Information retrieval; Database structures therefor; File system structures therefor of video data
- G06F16/73—Querying
- G06F16/735—Filtering based on additional data, e.g. user or group profiles
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/70—Information retrieval; Database structures therefor; File system structures therefor of video data
- G06F16/78—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually
- G06F16/7867—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually using information manually generated, e.g. tags, keywords, comments, title and artist information, manually generated time, location and usage information, user ratings
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Multimedia (AREA)
- Data Mining & Analysis (AREA)
- Databases & Information Systems (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Library & Information Science (AREA)
- Computational Linguistics (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention relates to a video recommendation method and device. The video recommendation method comprises the following steps: extracting character feature information from the meta-information of a target video; according to a user behavior log, sorting videos associated with the character feature information; and according to the sorted videos, displaying a recommendation result on the webpage of the target video. The embodiment of the invention extracts the character feature information from the meta-information of the target video and sorts the videos by combining with the user behavior log, the obtained recommended video has a high degree of association with the target video, and the influence on a recommendation result by character features can be reflected.
Description
Technical field
The present invention relates to multimedia technology field, particularly relate to a kind of video recommendation method and device.
Background technology
At present, when using Internet video to play the programs such as TV play, generally can show in webpage
Other content recommendation.For example, when playing TV play, essentially by collaborative filtering recommending method
Recommend UGC (User Generated Content, user's original content), or according to TV play title
Carry out mating recommending with the title of UGC video.
Use collaborative filtering recommending method, when playing a certain TV play, it is recommended that the video gone out mostly is same
The different collection of individual TV play, such as play the 10th collection, it is recommended that be the association content of the 2nd collection, it is recommended that go out
Video low with currently playing collection of drama relevance degree, it is recommended that the UGC number of videos gone out is fewer.
Summary of the invention
Technical problem
In view of this, the technical problem to be solved in the present invention is, how to recommend to close with target video to user
The video of Lian Dugao.
Solution
In order to solve above-mentioned technical problem, according to one embodiment of the invention, it is provided that a kind of video recommendations
Method, including:
Person characteristic information is extracted from the metamessage of target video；
According to User action log, the video associated with described person characteristic information is ranked up；
According to the video after sequence, the webpage of described target video shows recommendation results.
For said method, in a kind of possible implementation, extract from the metamessage of target video
Person characteristic information, including:
From described metamessage, obtain performer's sequence of described target video, and sort meter according to described performer
Calculate first weight of each performer of described target video；
From described metamessage, obtain the content introduction of described target video, and add up in described content introduction
The number of times that each role occurs；
According to the corresponding relation of performer in described metamessage Yu role, the number of times that each role occurs is converted to
The number of times that each performer occurs；
The number of times occurred according to each performer and the first weight, calculate performer's weight of described target video to
Amount.
For said method, in a kind of possible implementation, according to User action log, to institute
The video stating person characteristic information association is ranked up, including:
From described User action log, obtain each performer of described target video and associating of video
System；
From described User action log, obtain the search number of clicks peace of the video associated with each performer
All finish playing rate, and calculates the second weight of the video associated with each performer；
The video associated with each performer according to described second weight pair is ranked up, and obtains each performer corresponding
List of videos.
For said method, in a kind of possible implementation, according to User action log, to institute
The video stating person characteristic information association is ranked up, and also includes:
The comment data in setting the time period according to the filter word arranged and the video associated with each performer,
The list of videos that each performer is corresponding is adjusted.
For said method, in a kind of possible implementation, according to the video after sequence, described
Recommendation results is shown on the webpage of target video, including:
According to recommendation list length and described performer's weight vectors, determine the video recommendations that each performer is corresponding
Number；
The permutation and combination order of the video recommendations number corresponding according to each performer and setting, generates described recommendation and ties
Really.
According to another embodiment of the present invention, it is provided that a kind of video recommendations device, including:
Characteristic extracting module, for extracting person characteristic information from the metamessage of target video；
Order module, is connected with described characteristic extracting module, for according to User action log, to institute
The video stating person characteristic information association is ranked up；
Recommending module, for according to the video after sequence, showing recommendation on the webpage of described target video
Result.
For said apparatus, in a kind of possible implementation, described characteristic extracting module includes:
First weight calculation unit, for obtaining the performer row of described target video from described metamessage
Sequence, and first weight of each performer calculating described target video that sorts according to described performer；
Occurrence number statistic unit, is situated between for obtaining the content of described target video from described metamessage
Continue, and add up the number of times that in described content introduction, each role occurs；
Occurrence number converting unit, is connected with described occurrence number statistic unit, for according to described unit letter
In breath, performer and the corresponding relation of role, be converted to the secondary of each performer appearance by the number of times that each role occurs
Number；
Weight vector computation unit, with described first weight calculation unit and described occurrence number converting unit
Connect respectively, for the number of times occurred according to each performer and the first weight, calculate drilling of described target video
Member's weight vectors.
For said apparatus, in a kind of possible implementation, described order module includes:
Incidence relation acquiring unit, for from described User action log, obtains described target video
Each performer and the incidence relation of video；
Second weight calculation unit, for from described User action log, acquisition associates with each performer
The search number of clicks of video and the rate that averagely finishes playing, and calculate the second of the video associated with each performer
Weight；
List of videos sequencing unit, with described incidence relation acquiring unit and described second weight calculation unit
Connect respectively, be ranked up for the video associated with each performer according to described second weight pair, obtain each
The list of videos that performer is corresponding.
For said apparatus, in a kind of possible implementation, described order module also includes:
Filter element, is connected with described list of videos sequencing unit, for according to arrange filter word and with
The video of each performer association comment data in setting the time period, enters the list of videos that each performer is corresponding
Row sum-equal matrix.
For said apparatus, in a kind of possible implementation, described recommending module includes:
Recommend number to determine unit, for according to recommendation list length and described performer's weight vectors, determine each
The video recommendations number that performer is corresponding；
With described recommendation number, permutation and combination unit, determines that unit is connected, for according to corresponding the regarding of each performer
Frequency recommends the permutation and combination order of number and setting, generates described recommendation results.
Beneficial effect
The embodiment of the present invention, by extracting person characteristic information from the metamessage of target video, and combines
Video is ranked up by User action log, and the video of the recommendation obtained is high with the target video degree of association, energy
Enough reflection character features impacts on recommendation results.
According to below with reference to the accompanying drawings to detailed description of illustrative embodiments, the further feature of the present invention and side
Face will be clear from.
Accompanying drawing explanation
The accompanying drawing of the part comprising in the description and constituting description together illustrates with description
The exemplary embodiment of the present invention, feature and aspect, and for explaining the principle of the present invention.
Fig. 1 illustrates the flow chart of video recommendation method according to an embodiment of the invention；
Fig. 2 illustrates the flow process of characteristic extraction procedure in video recommendation method according to an embodiment of the invention
Figure；
Fig. 3 illustrates the flow chart of sequencer procedure in video recommendation method according to an embodiment of the invention；
Fig. 4 illustrates the flow chart of video recommendation method according to another embodiment of the present invention；
Fig. 5 illustrates the signal of TV play example in video recommendation method according to another embodiment of the present invention
Figure；
Fig. 6 illustrates the structural representation of video recommendations device according to an embodiment of the invention；
Fig. 7 illustrates the structural representation of video recommendations device according to another embodiment of the present invention.
Detailed description of the invention
Various exemplary embodiments, feature and the aspect of the present invention is described in detail below with reference to accompanying drawing.Attached
Reference identical in figure represents the same or analogous element of function.Although enforcement shown in the drawings
The various aspects of example, but unless otherwise indicated, it is not necessary to accompanying drawing drawn to scale.
The most special word " exemplary " means " as example, embodiment or illustrative ".Here as
Any embodiment illustrated by " exemplary " should not necessarily be construed as preferred or advantageous over other embodiments.
It addition, in order to better illustrate the present invention, detailed description of the invention below gives numerous
Detail.It will be appreciated by those skilled in the art that do not have some detail, the present invention is equally
Implement.In some instances, for method well known to those skilled in the art, means, element and circuit
It is not described in detail, in order to highlight the purport of the present invention.
Embodiment 1
Fig. 1 illustrates the flow chart of video recommendation method according to an embodiment of the invention.As it is shown in figure 1, should
Video recommendation method mainly may include that
Step 101, from the metamessage of target video extract person characteristic information；
Step 102, according to User action log, the video associated with described person characteristic information is carried out
Sequence；
Step 103, according to sequence after video, the webpage of described target video shows recommendation results.
Target video in the present embodiment refers to the video as subsequent recommendation foundation.The state of target video
Can be currently playing video, currently choose video etc. to be played.The type of target video can be
The collection of drama of TV play, it is also possible to be the video with metamessage such as film.Wherein, in the metamessage of video
Generally record this video character relation (such as: performer's sequence, the corresponding relation etc. of performer and role),
Content introduction etc..
In a kind of possible implementation, as in figure 2 it is shown, the characteristic extraction procedure of step 101 specifically wraps
Include:
Step 201, from the metamessage of target video, obtain performer's sequence of described target video, and root
Sort according to described performer and calculate first weight of each performer of described target video.
Wherein, the basic weight (the i.e. first weight) of performer can determine according to the sequence of performer.Mesh
The performer of mark video may have a lot, in order to alleviate computation burden, can only will act the leading role or part protagonist work
For calculating object.
Step 202, from described metamessage, obtain the content introduction of described target video, and add up described
The number of times that in content introduction, each role occurs.
Step 203, corresponding relation according to performer in described metamessage Yu role, occur each role
Number of times is converted to the number of times that each performer occurs.
Step 204, the number of times occurred according to each performer and the first weight, calculate drilling of described target video
Member's weight vectors.
Wherein, step 201 can be before step 201, it is also possible to after step 202, and the present invention is not
Limit the sequential relationship of step 201 and step 202.
In a kind of possible implementation, as it is shown on figure 3, the sequencer procedure of step 102 specifically includes:
Step 301, from described User action log, obtain each performer and the video of described target video
The incidence relation of (video to be recommended, such as UGC video).As a example by UGC video, from user behavior
Daily record can obtain user's click search behavior to UGC video, such as, be used for searching for certain performer
After, certain UGC video is clicked on and play.Within a period of time, searching for certain performer
Rope and UGC video are clicked on broadcasting behavior and are added up, it is possible to the UGC video obtaining associating with this performer
Search number of clicks and the rate that averagely finishes playing.
Step 302, from described User action log, obtain the Searching point of video associated with each performer
Hit number of times and the rate that averagely finishes playing, and calculate the second weight of the video associated with each performer.
Step 303, the video associated with each performer according to described second weight pair are ranked up, and obtain each
The list of videos that performer is corresponding.
In a kind of possible implementation, as it is shown on figure 3, the sequencer procedure of step 102 the most also includes:
Step 304, according to the filter word arranged and the video that associates with each performer in setting the time period
Comment data, is adjusted the list of videos that each performer is corresponding.
In a kind of possible implementation, as it is shown on figure 3, the recommendation process of step 103 specifically includes:
Step 305, according to recommendation list length and described performer's weight vectors, determine that each performer is corresponding
Video recommendations number.Generally, performer's weight vectors is the biggest, from the video that each performer that this performer is corresponding is corresponding
The video recommendations number chosen in list is the most.
Step 306, permutation and combination order according to video recommendations number corresponding to each performer and setting, generate
Described recommendation results.The permutation and combination order set can have multiple, and usual principle is the performer that weight is big
Before corresponding recommendation video comes, but certain the exposing of video that the little performer of weight to be given is corresponding
Chance, will intert and represent, and concrete example may refer to following embodiment 2.
The video recommendation method of the present embodiment, believes by extracting character features from the metamessage of target video
Breath, and combine User action log video is ranked up, the video of the recommendation obtained closes with target video
Lian Dugao, it is possible to the reflection character features impact on recommendation results.
Embodiment 2
Fig. 4 illustrates the flow chart of video recommendation method according to another embodiment of the present invention.As shown in Figure 4,
In the present embodiment, by target video be a certain integrate TV play, video to be recommended enters as a example by UGC video
Row exemplary illustration.Wherein, film or other have association metamessage video former with the recommendation of TV play
Reason is similar, and video to be recommended can also be other kinds of video, and citing is therefore not repeated.
Step 401, the feature of extraction TV play diversity video.
Step 4011, the basic weight of calculating TV play performer.
Specifically, can be according to the sequence (arranging by significance level) of performer, meter in TV play metamessage
Calculate the basic weight of performer.TV play metamessage generally can include the letter of each performers and clerks in TV play
Breath, content introduction etc..
Basic weight calculation formula is referred to formula 1:
Wherein, actorbaseweight is the basic weight of performer, and rank is the sequence of certain performer.Such as Fig. 5 institute
Show, as a example by TV play " biography discriminated by seraglio ", act the leading role and include " grandson pari, Chen Jianbin, Cai Shaofen, Liu
Setsuka, Li Tianzhu, Jiang Xin " etc..Can will act the leading role or more performer is as calculating object, it is also possible to
Select part performer as calculating object from act the leading role.The present embodiment use front three performer as calculating
Object.The result of calculation of the basic weight substituting into the available following each performer of equation 1 above that then performer sorted:
Grandson actorbaseweight pari=1；Actorbaseweight is old build refined=0.707；actorbaseweight
Cai Shaofen=0.577.
Step 4012, determine the corresponding relation of performer and role.
The corresponding relation of performer and role can be obtained from collection of drama metamessage.Connect example " biography discriminated by seraglio "
Corresponding relation is: Zhen-grandson pari；Emperor-Chen Jianbin；Queen-Cai Shaofen ...
Step 4013, from the diversity story of a play or opera that TV play often collects, extract role's title.
Specifically, the collection of drama brief introduction often collected according to TV play, extract the role in diversity collection of drama and occur
Number of times.Then, according to the data in step 4012, the number of times that role occurs is converted to performer and occurs
Number of times.
Such as, the synopsis of the 4th collection is as follows:
" Zhen stumbles on and has been buried greatly by people under the secret tree that Sui Yu pavilion cherry-apple tree do not blooms
Amount Moschus, this stays in the cause of fragrant high official abortion without reason of Sui Yu pavilion the most before this.Discriminate and fear that oneself obtains
Plotted to murder by other people, the most please Wen Shichu assists oneself, and false title is frightened to be caught wind and cold inconvenience and wait upon bedroom.Temperature
The most willingly emit capital of fence of beheading, disguise the truth for Zhen, and facilitate the Zhen state of an illness with medicine.Discriminate dye disease,
The plate in the emperor Bian Fanliaomei village, the gentle generous eyebrow village quite obtains emperor's favor.Next day, emperor grants a reward palace
In rare Lv Jugeimei village, and wish that she learns six palace matters, share somebody's cares and burdens for China prince wife.This measure allow China prince wife's anger
Fire burns.Meanwhile, queen blames the cowardliness of neat prince wife for, and another side the most extremely praises the eyebrow village, borrows
Machine expansile force.The cause of the death of maid of honour's good fortune finds out to be that China prince wife instigates other people to do through queen, dispatches a person to inform emperor
Supreme Being, the domineering behavior of emperor discontented China prince wife.At a time when now the quasi-Ge Er in northwest clan rebels, Man Chaowen
Have only in the middle of military officials of all ranks and descriptions before sending year thick soup Yao and go to put down a rebellion and can turn danger into safety.Emperor, for taking the interests of the whole into account, has to advise
Queen downplays the thing that good fortune is drowned.The disease discriminated never has improvement, the most sycophantic eunuch
Maids of honour seek another way out one after another.Head eunuch Kang Luhai is discontented with Zhen and loses power, and actively please resign the beautiful concubine of an emperor of service.
Discriminate exclamation to follow one's power or setback.”
Wherein, discriminating and occur in that 8 times, emperor occurs in that 6 times, and queen occurs in that 3 times
Then the number of times that above-mentioned role occurs being converted to performer's occurrence number is then:
Grandson actorfreq pari=8；Actorfreq is old build refined=6；Few sweet smell=3 of actorfreq Cai.
Step 4014, performer's weight of calculating diversity.
Assume that employing formula 2 calculates performer's weight of diversity:
Episode_actorweight=actorbaseweight × actorfreq formula 2,
Wherein, episode_actorweight is performer's weight, and actorbaseweight is the basic weight of performer,
Actorfreq is the number of times that performer occurs, then connecting example substitution formula 2 can obtain:
Grandson episode_actorweight pari=8.0；Episode_actorweight is old build refined=4.242；
Few sweet smell=1.731 of episode_actorweight Cai.
Then, use following formula 3 that above-mentioned performer's weight is normalized calculating:
Wherein, finalepisode_actorweight is diversity performer's weight vectors,
Episode_actorweight_sum is the sum of this collection all performers weight.
Connect example substitution formula 3 and then draw diversity performer's weight vectors.
Grandson finalepisode_actorweight pari=0.57, finalepisode_actorweight is old build refined
Few sweet smell=0.13 of=0.30, finalepisode_actorweight Cai.
Step 402, extraction UGC video features.
Step 4021, determine the UGC video associated with performer according to video search click logs.
According to some or the video search click logs of multiple video website such as youku every day, extract
UGC video and the incidence relation of performer.
In the such as Search Results of search " grandson pari ", corresponding UGC video, 00001 video clicks 100
Secondary, the rate that averagely finishes playing is 50%；00002 video search 80 times, the rate that averagely finishes playing is 90%；
00003 video search clicks 50 times, and the rate that averagely finishes playing is 20%.
Then can calculate, with reference to following formula 4, the UGC video weight associated with performer.
WeightUgc=clickfreq × averageplayrate formula 4,
Wherein, the UGC video weight that WeightUgc associates with a certain performer, for search, this drills clickfreq
Member also clicks on the number of times of certain UGC video, averageplayrate click on this UGC video after averagely broadcast
Putting completion rate, the rate that generally finishes playing is the ratio of reproduction time and video duration.
Connect example, then the UGC video and the weight that associate with grandson performer pari be:
WeightUgc00001=50.0；WeightUgc00002=72；WeightUgc00002=10.
The UGC video of grandson pari's association is ranked up as follows:
00002,00001,00003......
Step 4022, the UGC video of filtration title party.
For the UGC list of videos associated with performer, filter, it is possible to reduce the situation of title party.
For example, analyzing each UGC video (can be in the UGC list of videos associated with performer
Each UGC video is analyzed, it is also possible to be periodically analyzed the UGC video of whole video website)
The comment data of nearly a period of time (such as one month).When the comment data of a certain UGC video occurred
Filter word in filter dictionary, and occurrence number is more than when setting threshold value Nfreshold time, can and incite somebody to action this
UGC video marker is title party video.Then, deleting from the UGC list of videos associated with performer should
The relevant information of UGC video.Wherein it is possible to adjust threshold value Nfreshold according to actual needs.
Wherein, filter the word in dictionary can include " deceitful, title party, deceive click ... " etc. artificial
Or the filter word that computer sums up.
Step 403, generation TV play divide collection of drama recommendation list.
Assume the TV play a length of reclistLength=10 of diversity recommendation list.
Connect example, according to performer's weight vectors of the calculated diversity of equation 3 above be
Grandson finalepisode_actorweight pari=0.57, finalepisode_actorweight is old build refined=0.30,
Few sweet smell=0.13 of finalepisode_actorweight Cai.
Following formula 5 then can be used to calculate the UGC video counts associated in recommendation list with each performer:
Ugcnum=reclistLength × actorweight formula 5,
Wherein, ugcnum value can be with round number.Connect example and substitute into formula 5, then grandson ugcnum pari
=6, ugcnum be old build refined=3, ugcnum Cai few fragrant=1.
Assume in step 402, be calculated grandson pari and associate UGC list of videos and be
{A1,A2,A3,A4,A5,A6,A7....}；Chen Jianbin association UGC list is
{B1,B2,B3,B4,B5,B6.....}；Cai Shaofen association UGC list is { C1, C2, C3.....}.Then according to
The permutation and combination set can generate recommendation list.
For example, according to the performer's order in performer's weight vectors of diversity, insert one every time,
Recommendation list when eventually " seraglio discriminate pass " the 4th collection is play be A1, B1, C1, A2, B2, A3, B3, A4,
A5,A6}.It is of course also possible to insert two according to the performer's order in performer's weight vectors of diversity every time
(or multiple), such as { A1, A2, B1, B2, C1, A3, A4, B3, A5, A6}.Or, directly according to performer
Weight vectors order from high to low, in conjunction with the video recommendations number calculated, arranges UGC video, such as:
{A1,A2,A3,A4,A5,A6,B1,B2,B3,C1}。
In the present embodiment, carry out character features according to the video of TV play diversity collection of drama and extract and weight meter
Calculate, thus recommend according to diversity collection of drama performer's feature, it is possible to be acute according to the TV play difference play
Collection, it is recommended that the UGC video that content is different with emphasis is higher with the collection of drama degree of association play.
Embodiment 3
Fig. 6 illustrates the structural representation of video recommendations device according to an embodiment of the invention, such as Fig. 6 institute
Showing, this video recommendations device mainly may include that
Characteristic extracting module 61, for extracting person characteristic information from the metamessage of target video；
Order module 63, is connected with described characteristic extracting module 61, for according to User action log, right
The video associated with described person characteristic information is ranked up；
Recommending module 65, is connected with described order module 63, regarding after sorting according to order module 63
Frequently, the webpage of described target video shows recommendation results.
The video recommendation method that the video recommendations device of the present embodiment is able to carry out in above-described embodiment, passes through
From the metamessage of target video, extract person characteristic information, and combine User action log video is carried out
Sequence, the video of the recommendation obtained is high with the target video degree of association, it is possible to reflection character features is to recommending knot
The impact of fruit.
Embodiment 4
Fig. 7 illustrates the structural representation of video recommendations device according to another embodiment of the present invention.Fig. 7 gets the bid
Number assembly identical with Fig. 6 has identical function, does not repeats them here.
As it is shown in fig. 7, be with the difference of a upper embodiment, described characteristic extracting module 61 includes:
First weight calculation unit 611, for obtaining the performer of described target video from described metamessage
Sequence, and first weight of each performer calculating described target video that sorts according to described performer；
Occurrence number statistic unit 613, for obtaining the content of described target video from described metamessage
Introduce, and add up the number of times that in described content introduction, each role occurs；
Occurrence number converting unit 615, is connected with described occurrence number statistic unit 613, for according to institute
State performer and the corresponding relation of role in metamessage, the number of times that each role occurs is converted to each performer and occurs
Number of times；
Weight vector computation unit 617, turns with described first weight calculation unit 611 and described occurrence number
Change unit 615 to connect respectively, for the number of times occurred according to each performer and the first weight, calculate described mesh
Performer's weight vectors of mark video.
In a kind of possible implementation, described order module 63 includes:
Incidence relation acquiring unit 631, for from described User action log, obtains described target and regards
Each performer of frequency and the incidence relation of video；
Second weight calculation unit 633, for from described User action log, obtains and closes with each performer
The search number of clicks of video of connection and the rate that averagely finishes playing, and the video that associates with each performer of calculating
Second weight；
List of videos sequencing unit 635, with described incidence relation acquiring unit 631 and described second weight meter
Calculate unit 633 to connect respectively, arrange for the video associated with each performer according to described second weight pair
Sequence, obtains the list of videos that each performer is corresponding.
In a kind of possible implementation, described order module 63 also includes:
Filter element 637, is connected with described list of videos sequencing unit 655, for according to the filtration arranged
Word and the video that associates with each performer comment data in setting the time period, the video corresponding to each performer
List is adjusted.
In a kind of possible implementation, described recommending module 65 includes:
Number is recommended to determine unit 651, with weight vector computation unit 617 and list of videos sequencing unit 635
Connect respectively, for according to recommendation list length and described performer's weight vectors, determine that each performer is corresponding
Video recommendations number；Generally, performer's weight vectors is the biggest, from the video that each performer that this performer is corresponding is corresponding
The video recommendations number chosen in list is the most.Also, it is recommended to number determines that unit 651 can also be single with filtration
Unit 637 connects, and determines video recommendations number in the list of videos that each performer after filtration is corresponding.
With described recommendation number, permutation and combination unit 653, determines that unit 651 is connected, for according to each performer couple
The video recommendations number answered and the permutation and combination order of setting, generate described recommendation results.
The video recommendation method that the video recommendations device of the present embodiment is able to carry out in above-described embodiment, passes through
From the metamessage of target video, extract person characteristic information, and combine User action log video is carried out
Sequence, the video of the recommendation obtained is high with the target video degree of association, it is possible to reflection character features is to recommending knot
The impact of fruit.Concrete principle and example may refer to the associated description in above-mentioned video recommendation method.
The above, the only detailed description of the invention of the present invention, but protection scope of the present invention is not limited to
In this, any those familiar with the art, can be easily in the technical scope that the invention discloses
Expect change or replace, all should contain within protection scope of the present invention.Therefore, the protection of the present invention
Scope should be as the criterion with described scope of the claims.
Claims (10)
1. a video recommendation method, it is characterised in that including:
Person characteristic information is extracted from the metamessage of target video；
According to User action log, the video associated with described person characteristic information is ranked up；
According to the video after sequence, the webpage of described target video shows recommendation results.
Method the most according to claim 1, it is characterised in that carry from the metamessage of target video
Take person characteristic information, including:
From described metamessage, obtain performer's sequence of described target video, and sort meter according to described performer
Calculate first weight of each performer of described target video；
From described metamessage, obtain the content introduction of described target video, and add up in described content introduction
The number of times that each role occurs；
According to the corresponding relation of performer in described metamessage Yu role, the number of times that each role occurs is converted to
The number of times that each performer occurs；
The number of times occurred according to each performer and the first weight, calculate performer's weight of described target video to
Amount.
Method the most according to claim 2, it is characterised in that according to User action log, to
The video of described person characteristic information association is ranked up, including:
From described User action log, obtain each performer of described target video and associating of video
System；
From described User action log, obtain the search number of clicks peace of the video associated with each performer
All finish playing rate, and calculates the second weight of the video associated with each performer；
The video associated with each performer according to described second weight pair is ranked up, and obtains each performer corresponding
List of videos.
Method the most according to claim 3, it is characterised in that according to User action log, to
The video of described person characteristic information association is ranked up, and also includes:
The comment data in setting the time period according to the filter word arranged and the video associated with each performer,
The list of videos that each performer is corresponding is adjusted.
Method the most according to claim 4, it is characterised in that according to the video after sequence, in institute
State and on the webpage of target video, show recommendation results, including:
According to recommendation list length and described performer's weight vectors, determine the video recommendations that each performer is corresponding
Number；
The permutation and combination order of the video recommendations number corresponding according to each performer and setting, generates described recommendation and ties
Really.
6. a video recommendations device, it is characterised in that including:
Characteristic extracting module, for extracting person characteristic information from the metamessage of target video；
Order module, is connected with described characteristic extracting module, for according to User action log, to institute
The video stating person characteristic information association is ranked up；
Recommending module, for according to the video after sequence, showing recommendation on the webpage of described target video
Result.
Device the most according to claim 6, it is characterised in that described characteristic extracting module includes:
First weight calculation unit, for obtaining the performer row of described target video from described metamessage
Sequence, and first weight of each performer calculating described target video that sorts according to described performer；
Occurrence number statistic unit, is situated between for obtaining the content of described target video from described metamessage
Continue, and add up the number of times that in described content introduction, each role occurs；
Occurrence number converting unit, is connected with described occurrence number statistic unit, for according to described unit letter
In breath, performer and the corresponding relation of role, be converted to the secondary of each performer appearance by the number of times that each role occurs
Number；
Weight vector computation unit, with described first weight calculation unit and described occurrence number converting unit
Connect respectively, for the number of times occurred according to each performer and the first weight, calculate drilling of described target video
Member's weight vectors.
Device the most according to claim 7, it is characterised in that described order module includes:
Incidence relation acquiring unit, for from described User action log, obtains described target video
Each performer and the incidence relation of video；
Second weight calculation unit, for from described User action log, acquisition associates with each performer
The search number of clicks of video and the rate that averagely finishes playing, and calculate the second of the video associated with each performer
Weight；
List of videos sequencing unit, with described incidence relation acquiring unit and described second weight calculation unit
Connect respectively, be ranked up for the video associated with each performer according to described second weight pair, obtain each
The list of videos that performer is corresponding.
Device the most according to claim 8, it is characterised in that described order module also includes:
Filter element, is connected with described list of videos sequencing unit, for according to arrange filter word and with
The video of each performer association comment data in setting the time period, enters the list of videos that each performer is corresponding
Row sum-equal matrix.
Device the most according to claim 9, it is characterised in that described recommending module includes:
Recommend number to determine unit, for according to recommendation list length and described performer's weight vectors, determine each
The video recommendations number that performer is corresponding；
With described recommendation number, permutation and combination unit, determines that unit is connected, for according to corresponding the regarding of each performer
Frequency recommends the permutation and combination order of number and setting, generates described recommendation results.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610151232.6A CN105843857B (en) | 2016-03-16 | 2016-03-16 | Video recommendation method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610151232.6A CN105843857B (en) | 2016-03-16 | 2016-03-16 | Video recommendation method and device |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN105843857A true CN105843857A (en) | 2016-08-10 |
| CN105843857B CN105843857B (en) | 2019-08-27 |
Family
ID=56587092
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610151232.6A Active CN105843857B (en) | 2016-03-16 | 2016-03-16 | Video recommendation method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105843857B (en) |
Cited By (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106407401A (en) * | 2016-09-21 | 2017-02-15 | 乐视控股（北京）有限公司 | A video recommendation method and device |
| CN107122462A (en) * | 2017-04-27 | 2017-09-01 | 深圳市茁壮网络股份有限公司 | The method and apparatus that incidence relation between each matchmaker's money program is set up based on character data |
| CN108134947A (en) * | 2017-12-25 | 2018-06-08 | 横琴国际知识产权交易中心有限公司 | A kind of video intercepting method and device |
| CN108471544A (en) * | 2018-03-28 | 2018-08-31 | 北京奇艺世纪科技有限公司 | A kind of structure video user portrait method and device |
| CN109446419A (en) * | 2018-10-17 | 2019-03-08 | 武汉斗鱼网络科技有限公司 | A kind of method and device for recommending video |
| CN110858914A (en) * | 2018-08-23 | 2020-03-03 | 北京优酷科技有限公司 | Video material recommendation method and device |
| CN112967086A (en) * | 2021-02-27 | 2021-06-15 | 葛纪侠 | Intelligent marketing promotion method and device and electronic equipment |
| CN113055707A (en) * | 2019-12-26 | 2021-06-29 | 青岛海信传媒网络技术有限公司 | Video display method and device |
| CN114615525A (en) * | 2022-03-18 | 2022-06-10 | 北京字节跳动网络技术有限公司 | Video pushing method, device, equipment and storage medium |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102685566A (en) * | 2012-05-28 | 2012-09-19 | 北京网尚数字电影院线有限公司 | Recommendation method for audio and video programs |
| US8452797B1 (en) * | 2011-03-09 | 2013-05-28 | Amazon Technologies, Inc. | Personalized recommendations based on item usage |
| CN104199896A (en) * | 2014-08-26 | 2014-12-10 | 海信集团有限公司 | Video similarity determining method and video recommendation method based on feature classification |
| CN104216960A (en) * | 2014-08-21 | 2014-12-17 | 北京奇艺世纪科技有限公司 | Method and device for recommending video |
| CN104219575A (en) * | 2013-05-29 | 2014-12-17 | 酷盛(天津）科技有限公司 | Related video recommending method and system |
-
2016
- 2016-03-16 CN CN201610151232.6A patent/CN105843857B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8452797B1 (en) * | 2011-03-09 | 2013-05-28 | Amazon Technologies, Inc. | Personalized recommendations based on item usage |
| CN102685566A (en) * | 2012-05-28 | 2012-09-19 | 北京网尚数字电影院线有限公司 | Recommendation method for audio and video programs |
| CN104219575A (en) * | 2013-05-29 | 2014-12-17 | 酷盛(天津）科技有限公司 | Related video recommending method and system |
| CN104216960A (en) * | 2014-08-21 | 2014-12-17 | 北京奇艺世纪科技有限公司 | Method and device for recommending video |
| CN104199896A (en) * | 2014-08-26 | 2014-12-10 | 海信集团有限公司 | Video similarity determining method and video recommendation method based on feature classification |
Non-Patent Citations (1)
| Title |
|---|
| 周娇等: "社交网络服务中的多维空间视频推荐算法", 《计算机工程》 * |
Cited By (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106407401A (en) * | 2016-09-21 | 2017-02-15 | 乐视控股（北京）有限公司 | A video recommendation method and device |
| CN107122462A (en) * | 2017-04-27 | 2017-09-01 | 深圳市茁壮网络股份有限公司 | The method and apparatus that incidence relation between each matchmaker's money program is set up based on character data |
| CN108134947A (en) * | 2017-12-25 | 2018-06-08 | 横琴国际知识产权交易中心有限公司 | A kind of video intercepting method and device |
| CN108471544B (en) * | 2018-03-28 | 2020-09-15 | 北京奇艺世纪科技有限公司 | Method and device for constructing video user portrait |
| CN108471544A (en) * | 2018-03-28 | 2018-08-31 | 北京奇艺世纪科技有限公司 | A kind of structure video user portrait method and device |
| CN110858914A (en) * | 2018-08-23 | 2020-03-03 | 北京优酷科技有限公司 | Video material recommendation method and device |
| CN110858914B (en) * | 2018-08-23 | 2021-11-26 | 阿里巴巴（中国）有限公司 | Video material recommendation method and device |
| CN109446419A (en) * | 2018-10-17 | 2019-03-08 | 武汉斗鱼网络科技有限公司 | A kind of method and device for recommending video |
| CN113055707A (en) * | 2019-12-26 | 2021-06-29 | 青岛海信传媒网络技术有限公司 | Video display method and device |
| CN113055707B (en) * | 2019-12-26 | 2023-07-11 | 青岛海信传媒网络技术有限公司 | Video display method and device |
| CN112967086A (en) * | 2021-02-27 | 2021-06-15 | 葛纪侠 | Intelligent marketing promotion method and device and electronic equipment |
| CN112967086B (en) * | 2021-02-27 | 2022-08-02 | 北京橙色风暴数字技术有限公司 | Intelligent marketing promotion method and device and electronic equipment |
| CN114615525A (en) * | 2022-03-18 | 2022-06-10 | 北京字节跳动网络技术有限公司 | Video pushing method, device, equipment and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN105843857B (en) | 2019-08-27 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105843857A (en) | Video recommendation method and device | |
| Weinstein et al. | Persistent bill and corolla matching despite shifting temporal resources in tropical hummingbird‐plant interactions | |
| CN102737029B (en) | Searching method and system | |
| CN110704674B (en) | Video playing integrity prediction method and device | |
| Finin et al. | Annotating named entities in Twitter data with crowdsourcing | |
| CN107633084A (en) | Based on the public sentiment managing and control system and its method from media | |
| CN104769957B (en) | Identification and the method and apparatus that internet accessible content is presented | |
| CN104462553B (en) | Question and answer page relevant issues recommended method and device | |
| Schoeffmann | A user-centric media retrieval competition: The video browser showdown 2012-2014 | |
| CN110737859B (en) | UP master matching method and device | |
| CN103139616A (en) | Personalization cloud recommendation method and system based on multi-user digital television program | |
| CN104219575A (en) | Related video recommending method and system | |
| Hanley et al. | No calm in the storm: investigating QAnon website relationships | |
| Singh et al. | Mimoqa: Multimodal input multimodal output question answering | |
| CN110895568B (en) | Method and system for processing court trial records | |
| CN107846611A (en) | A kind of TV programme method for pushing and system based on age bracket | |
| CN106355446A (en) | Online and mobile game advertising recommending system | |
| CN109033072A (en) | A kind of audiovisual material supervisory systems Internet-based | |
| CN108280124A (en) | Product classification method and device, ranking list generation method and device, electronic equipment | |
| Blaszka et al. | Was dissent being displayed during the Sochi Olympics? Examining the# Sochi2014 hashtag for dissent | |
| Jalal et al. | Scraping social media photos posted in Kenya and elsewhere to detect and analyze food types | |
| Goëau et al. | Pl@ ntnet mobile 2014: Android port and new features | |
| KR101269205B1 (en) | Multilanguage information offer system of one - website | |
| CN109918525A (en) | Food picture esthetic analysis label data collection system based on wechat small routine | |
| Teixeira-Costa et al. | Comparative phenology of mistletoes shows effect of different host species and temporal niche partitioning |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| CP03 | Change of name, title or address | ||
| CP03 | Change of name, title or address |
Address after: 100080 Beijing Haidian District city Haidian street A Sinosteel International Plaza No. 8 block 5 layer A, C Patentee after: Youku network technology (Beijing) Co.,Ltd. Address before: 100080 area a and C, 5 / F, block a, Sinosteel International Plaza, No. 8, Haidian Street, Haidian District, Beijing Patentee before: 1VERGE INTERNET TECHNOLOGY (BEIJING) Co.,Ltd. |
|
| TR01 | Transfer of patent right | ||
| TR01 | Transfer of patent right |
Effective date of registration: 20200526 Address after: 310052 room 508, floor 5, building 4, No. 699, Wangshang Road, Changhe street, Binjiang District, Hangzhou City, Zhejiang Province Patentee after: Alibaba (China) Co.,Ltd. Address before: 100080 Beijing Haidian District city Haidian street A Sinosteel International Plaza No. 8 block 5 layer A, C Patentee before: Youku network technology (Beijing) Co.,Ltd. |